Abstract
Passive safety systems, which do not need external input to operate, are widely used to enhance the inherent safety. They are receiving increasing attention due to the remarkable advantages in the simplicity and reduction of human interactions. The main task of passive system reliability assessment lies in the estimation of failure probability taking various uncertainties into account. However, it is usually a challenging work based on the traditional Monte-Carlo method because repeatedly running thermal-hydraulics code (e.g., RELAP5) is required. Unfortunately, it is computationally impractical because a large number of simulations are needed. In addition, each simulation of such thermal-hydraulic code may take hours, the computational cost of which is prohibitively high. In order to address this problem, in the present work we propose an adaptive sampling method for Kriging metamodel. The proposed approach selects samples strategically based on the information from the previous iteration. The proposed method is tested with reference to one benchmark case and then applied to the reliability assessment of the passive residual heat removal system in an integral type PWR (IPWR200). Results indicated that this adaptive method decreases the calls of performance function to construct a Kriging model and improves the calculational efficiency to a great extent.
